Book Appointment

A Ninety-Day IAMOps Access Audit Checklist for Clinic Back Offices

Healthcare Software

Imagine a facilities coordinator at a mid-sized orthopaedic group, a character I am inventing for this walk-through. One afternoon she tries to find out who can open the shared leaflet folder on the organisation’s file system. Not patient records, just the printable forms the website team edits. It takes three people and two days, and the answer surprises everyone. That small embarrassment is a good reason to run an access audit, and the checklist below is how I would do it. It follows the thinking behind IAMOps, and it deals strictly with administrative systems, never with clinical care.

Why a checklist rather than a project plan

Checklists work because they survive busy weeks. A project plan lives in somebody’s head until the person moves on; a list on a shared page can be picked up by anyone. The Ops Atlas sketches a three-stage arc of about a month each in its write-up of the IAMOps practice, and the audit below borrows that rhythm while keeping the wording and the steps my own.

Days 1 to 30: find out what exists

Start with plain curiosity. You are not fixing anything yet.

  • Name an owner for every system. If you cannot, write “unknown” and treat that as the first finding.
  • List who and what can enter. Staff, contractors, automated jobs, third-party tools connected through sign-in buttons.
  • Look at outside reach. Which files, buckets or pages are visible to people outside your organisation? Several cloud platforms provide a free analysis for exactly this.
  • Pick one critical automated pipeline and switch it from a stored secret to a short-lived credential.
  • Write one protective rule and test it in a sandbox before it touches anything real.

Our imagined coordinator finishes this month with a list of forty-odd accounts and a surprising discovery: a former agency still had editing rights on the leaflet folder. Nobody had done anything wrong. The access had simply outlived the contract, which is how access usually ends up lingering. A short note in the register records what was found and who agreed to close it, so the discovery does not evaporate by the next meeting.

Days 31 to 60: put the rules somewhere you can review them

  • Create a repository for policies and move the current ones in, with owners listed in the file header.
  • Add an automated check on every proposed change so reviewers see problems before approving.
  • Introduce one delegated limit. Let a team manage its own rules but within a ceiling set above them, so freedom comes with a boundary.
  • Run an unused-access report and start with the easy cases: permissions no one has exercised in a long time.
  • Record baselines. You cannot show improvement without a starting figure.

A note on container platforms

If your IT group runs Kubernetes for any of its web properties, remember two facts the Atlas stresses. Its permission system is purely additive, so a rule can add access but never subtract it; guardrails belong in admission control instead. And the right to list the secrets in a namespace actually exposes their contents, so treat it as read access, not as a harmless directory view.

Days 61 to 90: make it a habit

  • Federate half of the remaining pipelines so fewer secrets live in settings pages.
  • Hold the first monthly review, even if it lasts twenty minutes.
  • Open an exception register with owner, reason and review date for every deviation.
  • Decide on a policy engine or consciously decide not to yet, and write down why.

By this point the coordinator, who never thought of herself as technical, runs the monthly review herself. To her it feels like a fire-door inspection: unglamorous, and the reason she sleeps better.

The audit questions, condensed

  1. Who owns it?
  2. Who and what can reach it?
  3. Can anyone outside reach it?
  4. When did we last see this permission used?
  5. Is the credential short-lived?
  6. Is the rule in version control and tested?
  7. If we removed it today, who would shout first?
  8. When will we look again?

The seventh question sounds flippant yet is the most useful. If the answer is “nobody”, you have probably found something to remove. If it names a person, you have found an owner and a conversation to have.

Traps to watch for while you audit

Allow statements that use a “not action” form can grant more than their author intended, because they permit everything except the listed items, including actions added to the platform later. Rules that exempt a particular role by its name need a companion rule that stops anyone from creating or editing something with that name. And remember that the Atlas lists several well-known incidents, from a 2021 court case about exposed role credentials to a 2024 email-system compromise through a forgotten test application, as reminders that old, quiet permissions are what attackers look for first.

What to do with the findings

Rank them by how much damage a mistake could do and how easy the fix is. Fix the cheap, risky ones at once. Schedule the expensive ones with an owner. Record the ones you accept, with a reason. An audit that ends in a tidy report and no changes is a pleasant document, but it is not an improvement.

Share progress in plain numbers where you can. A count of accounts without owners, a count of keys past three months old, and a count of exceptions overdue for review fit on a single slide and need no jargon. When the counts fall from one quarter to the next, colleagues who never attend the technical meetings can still see that the effort is paying off, and they become allies instead of bystanders.

If you would like extra reading on how teams run automated systems day to day, there is a practical look at managing growing fleets of working robots and an architectural take on joined-up IT operations. Both frame the same habit: routine checks beat heroic rescues.

In the story’s last scene the coordinator keeps the audit list pinned to the shared drive. Every quarter she changes the date at the top and walks the eight questions again. It takes an afternoon, and the leaflet folder never surprises anybody again.

Real organisations will find their own surprises, and they will rarely be dramatic. More often they are ordinary leftovers: a shared login that three people still use, a test integration from a long-finished project, a contractor group nobody remembered to close. The audit is valuable precisely because it turns those leftovers from vague worries into named, ranked, assignable items. Once something has a name and an owner, it tends to get fixed.